In this study, the pCAMBIA1302 vector was used for generating transgenic rice lines. The coding sequence of OsATG8a (Os07g0512200; 360 bp in length) was amplified from rice cDNA and inserted at the Eco91I site through homologous recombination as described (Luo et al., 2017). To generate ProUBQ10:GFP-OsATG8a and ProATG8a:GFP-OsATG8a plasmids, the 35S promoter was replaced with the ubiquitin promoter and the OsATG8a promoter, respectively, and inserted between HindIII and NcoI sites. All constructs were verified by DNA sequencing. The primers used for cloning, plasmid construction, and sequencing are listed in Supplementary Table S1.

Transgenic rice lines were generated similarly to a previous report on OsATG8s (Izumi et al., 2015). Briefly, sterilized rice seeds (japonica rice cultivar Jinjing 818) were used for callus induction. The vectors were transformed into mature seed-derived rice calli by Agrobacterium-mediated transformation. Transgenic rice plants (T1) were obtained through several rounds of differentiation inductions and selected on medium containing Hygromycin B. These T1 transgenic rice lines were selected by liquid medium containing 50 mg/L Hygromycin B for 3 days. After germinating in plates containing water for 5–6 days, the seedlings were grown in the modified Hoagland’s solution (5 mM KNO₃, 2 mM MgSO₄·7H₂O, 5 mM Ca(NO₃)₂·4H₂O, 1 mM KH₂PO₄, 0.05 mM FeSO₄·7H₂O, 0.05 mM Na₂EDTA, 46 nM H₂BO₃, 9 nM MnCl₂·4H₂O, 0.3 nM CuSO₄·5H₂O, and 0.8 nM ZnSO₄·7H₂O) at 28°C/24°C, 14 h light/10 h dark in a growth chamber. Transgenic rice was cultivated in paddy fields in growth seasons of 2019 to 2021 in Tianjin, Jiangsu, and Hainan provinces of China.

GFP-OsATG8a driven by three different promoters were transiently expressed in tobacco leaves as described (Luo et al., 2017). Soil grown, 4-week-old N. benthamiana leaves were used for Agrobacterium infiltration. After 2 days of inoculation, leaves were collected and cut into small squares for confocal microscopy.

Sterilized rice seeds (japonica rice cultivar Jinjing 818) were germinated on 1/2 Murashige and Skoog (MS) medium with a photoperiod of 14 h light and 10 h dark at 26°C for 5–6 days, then moved to the dark for another 5–6 days. The etiolated stem and sheath of rice seedlings were cut into pieces of approximately 0.5 mm with sharp razors. These pieces were immediately transferred into 20 ml enzyme solution (1.5% Cellulase RS, 0.75% Macerozyme R-10, 0.6 M mannitol, 10 mM MES at pH 5.7, 10 mM CaCl₂, and 0.1% BSA), vacuum infiltrated for 30 min, and further digested in the dark for 4–5 h with gentle shaking. After enzyme digestion, protoplasts were released by adding an equal volume of W5 solution (154 mM NaCl, 125 mM CaCl₂, 5 mM KCl, and 2 mM MES at pH 5.7) and gentle shaking by hand for 2 min. Then, the protoplasts were filtered through a 40 μm-gauge nylon mesh with 3–5 washes using W5 solution and collected by spinning at 100 g for 3 min. After protoplasts were re-suspended and washed once with W5 solution, MMG solution (0.4 M mannitol, 15 mM MgCl₂, and 4 mM MES at pH 5.7) was used to re-suspend the pellets at a concentration of 2 × 10⁶ cells ml⁻¹. For each transformation, 6–8 μg of freshly prepared plasmid DNA and 200 μl protoplasts (about 4 × 10⁵ cells) were mixed with 220 μl freshly prepared PEG solution [40% (w/v) PEG 4000, 0.4 M mannitol, and 0.1 M CaCl₂] and were incubated at room temperature for 30 min in the dark. After incubation, the protoplasts were mixed with 1 ml W5 solution and incubated at 28°C for 12–15 h in the dark before imaging.

Transiently transformed tobacco leaves (lower epidermis) and rice protoplasts were observed with a Ni-E A1 HD25 confocal microscope (Nikon, Japan). Prior to image collection, the background auto-fluorescence was eliminated using untransformed tobacco leaves or rice protoplasts. The GFP fluorescence signal was exited at 488 nm and emission was collected at 500–550 nm. The chlorophyll auto-fluorescence was exited with 561 nm laser and emission was collected at 600–700 nm. For NaCl treatments, protoplasts were incubated in W5 solution with 250 mM NaCl, or 50 μM E-64d, or both NaCl and E-64d, for 30 min at room temperature. For inhibitor treatments, transformed protoplasts were incubated in W5 solution containing 200 nM AZD8055, or 1 μM ConA, or both AZD8055 and ConA, for 3 h at room temperature.

Leaves from 14-day-old seedlings were used for RNA extraction, cDNA synthesis, and quantitative real-time RT-PCR (qRT-PCR) as described (Luo et al., 2017). Transcript levels of GFP and OsATG8a were normalized to OsEF1a (LOC_Os03g08020) with three biological replicates consisting of four technical repeats. Specific amplification was verified by a melt curve analysis following the completion of the PCR cycles. Each PCR product generated a single peak in melt curve analysis, indicating specific amplification. The 2^-ΔΔCT method was used for relative quantification of qRT-PCR data. Primers used are listed in Supplementary Table S1.

Two-week-old seedlings were used for autophagic flux measurement. Excised leaves were incubated in the Hoagland’s solution with 0.01% Tween-20, and stirred and vacuumed to make sure the leaves were completely soaked in the solution. For NaCl treatments, excised leaves were incubated in the Hoagland’s solution plus 150 mM NaCl, or 100 μM E-64d, or both NaCl and E-64d, for 1 h at room temperature. Alternatively, 250 mM NaCl was used. For inhibitor treatments, excised leaves were incubated in the Hoagland’s solution containing 2 μM AZD8055, or 1 μM ConA, or both AZD8055 and ConA, for 6 h at room temperature. In addition, 6 h with 200 μM BTH plus 100 μM E-64d treatment and 4 h with 2 mM DTT plus 100 μM E-64d treatment were used to induce autophagy. Protein extraction and immunoblotting were done as described (Luo et al., 2017). Semi-quantification of the protein levels was performed with ImageJ¹ and protein levels were normalized to Coomassie brilliant blue R-250-stained band of the RuBisCO large subunit (RbcL). For immunoblotting, mouse anti-GFP (1:5,000 dilution, Utibody, China), rabbit anti-H3 (1:8,000 dilution, ABclonal, China), and the appropriate IgG-HRP conjugated secondary antibody (1:5,000; ZSGB-Bio, China) were used. The signal was developed using High-sensitivity ECL chemiluminescence detection kit (Vazyme, China) and chemiluminescence was detected using a chemiluminescent Western Blot scanner (ChemiScope 6100T, Clinx, China). All experiments were repeated for three to five times, and one representative result was shown.

Transgenic rice seedlings were frozen and ground in liquid nitrogen, and protein extraction buffer (50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 20% Glycerol, 1 mM MgCl₂, 0.2% NP-40, and 1% Protease Inhibitor Cocktail) was added at a proportion of 1:2 (m:v). Extracts were centrifuged at 12,000 rpm for 15 min at 4°C. Then, the supernatants were collected and centrifuged again at 12,000 rpm for 5 min at 4°C. The supernatants (1 ml to 10 μl bead volume) were incubated with GFP-Trap A beads pre-equilibrated in wash buffer (50 mM Tris–HCl, pH 7.5, 150 mM NaCl, 20% Glycerol, 1 mM MgCl₂, and 0.1% NP-40) at 4°C for 2–3 h. The beads were washed 3–4 times with wash buffer and were re-suspended in 100 μl wash buffer. Samples were verified by Western blotting; the corresponding gels were cut and sent for mass spectrometry analysis at the Instrumental Analysis Center of Shanghai Jiao Tong University.

We cloned OsATG8s from an herbicide-resistant japonica rice cultivar Jinjing 818 and examined their transcript levels. Consistent with previous reports (Xia et al., 2011; Izumi et al., 2015), OsATG8a was the most highly expressed isoform. Hence, it was selected for vector construction (Figure 1A). OsATG8d was also selected (Supplementary Figure S2); however, we were unable to obtain transgenic lines carrying ProATG8d:GFP-OsATG8d, and it was eventually left out.

Expression of GFP-OsATG8a driven by 35S, ubiquitin, and OsATG8a (endogenous) promoters was firstly examined in transiently transformed N. benthamiana leaf epidermal cells. Pro35S:GFP gave typical, strong nuclear, and cytoplasmic GFP signals. Pro35S:GFP-OsATG8a gave relatively weak cytoplasmic signals and punctate signals that likely represent autophagosomes. Strong cytoplasmic/endoplasmic reticulum and punctate signals were observed from cells transformed with ProUBQ10:GFP-OsATG8a (Figure 1B). ProATG8a:GFP-OsATG8a had the lowest expression level, with puncta representing autophagosomes observed (Figure 1B).

We then examined the expression levels of the three constructs in transiently transformed rice protoplasts (Figure 1C). Again, ProUBQ10:GFP gave cytoplasmic signals, whereas the GFP-ATG8a constructs gave both cytoplasmic and punctate signals. Similar to our observation in tobacco leaves, ProUBQ10:GFP-ATG8a gave strong and clear signals, and signal from GFP-ATG8a driven by the endogenous promoter was very weak.

We had previously shown that the autophagic flux can be induced by salt stress within 30 min in Arabidopsis (Luo et al., 2017). Here, we also treated the transformed rice protoplasts with 250 mM NaCl in the presence of a lysosomal cysteine protease inhibitor E-64d to see if autophagic flux could be induced by salt treatment (Figure 2). GFP-OsATG8a driven by all three promoters responded to salt treatment, as NaCl plus E-64d significantly induced GFP-OsATG8a accumulation (Figures 2A–C).

Since inhibition of Target of Rapamycin (TOR) can reliably induce autophagy, we treated the protoplasts with 200 nM AZD8055, a potent TOR inhibitor (Chresta et al., 2010), in the presence of the V-ATPase inhibitor ConA (Huss et al., 2002), to further validate the constructs. GFP-OsATG8a driven by UBQ10 and ATG8a promoters both responded to TOR inhibition (Supplementary Figure S1).

Inferring from the expression intensities, the ubiquitin promoter is a good choice for the transient expression of OsATG8a.

After validating and comparing the three constructs in transient expression systems, transgenic rice carrying these constructs were generated. T1 transgenic lines regenerated from the transformed calli were firstly validated with genomic PCR and immunoblotting (Supplementary Figures S2A,B). Then, the plants were grown in the paddy fields and measured for their heights and the number of tillers before harvesting. Consistent with a recent report (Fan et al., 2020), the transgenic lines were not very different among themselves in their heights (Supplementary Figure S3A). The numbers of tillers were not very different either (Supplementary Figure S3B). No significant change in phenotype was observed at the seedling stage in the T3 transgenic lines (Supplementary Figure S3C).

The T3 transgenic seedlings carrying Pro35S:GFP-OsATG8a, ProUBQ10:GFP-OsATG8a, and ProOsATG8a:GFP-OsATG8a were evaluated for their suitability in measuring autophagic flux. The autophagic flux was measured with the GFP-ATG8 cleavage assay, in which the flux is represented by the ratio of the amount of cleaved GFP (27 kD) to total GFP in the lane (40 kD plus 27 kD), i.e., GFP/(GFP + GFP-ATG8a). Firstly, GFP-OsATG8a (40kD) and free GFP (27kD) bands were readily detected in the transgenic lines (Figure 3). Since one or two bands were detected for GFP-OsATG8a on immunoblots, we validated the bands by running the samples along with recombinant GFP-OsATG8s purified from E.coli. Two bands of similar sizes were seen for the recombinant proteins GFP-OsATG8a and GFP-OsATG8d (Supplementary Figure S2C). Such observation indicated that both bands correspond to GFP-ATG8, and the amount of GFP-ATG8 should be measured as the sum of the two.

For autophagy induction, rice seedlings were treated with NaCl, AZD8055, BTH, and DTT (Figures 3A–E). Benzo-(1,2,3)-thiadiazole-7-carbothioic acid (BTH), a salicylic acid agonist, has been used to induce autophagy (Yoshimoto et al., 2009; Zeng et al., 2021). DTT is known to induce ER stress-elicited autophagy (Liu et al., 2012). In general, GFP-OsATG8a driven by all three promoters responded mildly to the treatments and chemicals. A possible explanation is that the basal level autophagy is relatively high already in rice. Specifically, Pro35S:GFP-OsATG8a responded relatively strongly to 200 μM BTH. ProUBQ10:GFP-OsATG8a responded well to 250 mM NaCl and BTH, but not to 2 mM DTT. ProOsATG8a:GFP-OsATG8a responded nicely to moderate and severe salt stress, AZD8055, and DTT. A diagram illustrating the GFP cleavage assay and the chemicals used is presented (Figure 3F). E-64d is omitted from the diagram because its precise role in preserving GFP, or even GFP-ATG8, is currently unknown.

At this point, we noticed an interesting phenomenon. The T3 transgenic lines carrying 35S:GFP and Ubiquitin:GFP gave strong GFP bands (Figure 4A; Supplementary Figure S2B), validating the strength of the promoters. However, lines carrying GFP-OsATG8a driven by 35S, ubiquitin, or OsATG8a promoters had comparable protein levels of GFP-OsATG8a among themselves (Figure 4A, Supplementary Figure S2B). We then examined the transcript levels of the GFP-OsATG8a fusion gene and the endogenous OsATG8a with qRT-PCR (Figures 4B–E). To distinguish ectopic OsATG8a from the endogenous transcript, two pairs of primers were used for OsATG8a, with one targeting the coding sequence and the other targeting the 3’-UTR (Figure 4B). Consistent with the protein levels (Figure 4A), when expressed alone, the GFP transcript can accumulate to high levels if driven by 35S or ubiquitin promoters (Figure 4C). In contrast, when expressed as GFP-OsATG8a, the transcript level of GFP, representing the fusion gene, was clearly repressed (Figure 4C), indicative of transcriptional regulation. OsATG8a transcripts, representing both endogenous and ectopic OsATG8a, accumulated in all GFP-ATG8s lines carrying the three constructs, however not very differently except for one line (Figure 4D), yet again suggesting a regulation at the transcription level. Judging from the levels of the OsATG8a-UTR, the endogenous expression of OsATG8a was not very different in all lines including the wild type, suggesting that it is the transcription of the fusion gene that gets regulated (Figure 4E). Therefore, both transcriptional and post-transcriptional regulation of ATG8a have taken place in the transgenic lines carrying GFP-ATG8a, especially in Pro35S:GFP-OsATG8a and ProUBQ10:GFP-OsATG8a, resulting in comparable GFP-OsATG8a protein levels. The mechanism is currently unknown and awaits further study.

ATG8 is a star molecule in proteomic studies of autophagy. To see if the GFP-OsATG8a transgenic lines could be used for protein–protein interaction studies, GFP-Trap immunoprecipitation was performed using seedlings from three transgenic lines along with GFP control plants grown under control conditions (mock = Hoagland’s solution) or treated with Hoagland’s solution containing 150 mM NaCl for 1 h. The immuno-precipitated proteins were verified with SDS-PAGE and sent for mass spectrometry analyses.

From the eight conditions, 2,244 proteins from 8,120 peptides altogether were obtained. After subtracting the proteins identified in GFP control, 56, 97, and 86 proteins remained for Pro35S:GFP-OsATG8a, ProUBQ10:GFP-OsATG8a, and ProOsATG8a:GFP-OsATG8a under control conditions (Figure 5A). With salt treatment, 54, 106, and 97 proteins were identified as interacting partners for OsATG8a in the three lines, respectively (Figure 5B). The numbers of shared target proteins between different transgenic lines were also shown in the Venn diagrams (Figures 5A,B). To better describe the OsATG8a-interacting protein landscape, we mapped the proteins to the Arabidopsis proteome by batch BLAST and used the Arabidopsis protein with the lowest e-value to represent the rice protein. Then, we combined all target proteins identified from mock or NaCl treatment for Gene Ontology (GO) analysis using Cytoscape as described (Wu et al., 2022). Under controlled (mock) conditions, most enriched GO terms are tryptophan metabolic processes, indolalkylamine metabolic processes, photosynthesis, and starch biosynthesis, indicative of growth-related processes like auxin biosynthesis and photosynthesis (Figure 5C). Such categorization is consistent with the knowledge that basal level autophagy constitutively recycles damaged and obsolete proteins. With salt stress, GO terms related to the amino acid metabolic processes, especially serine, aspartate, and cysteine metabolic processes overwhelmingly enriched (Figure 5D). Energy reserve and starch metabolic processes were also enriched (Figure 5D).

We also randomly generated a list of cytoplasmic proteins that have not been reported to interact with ATG8 (Supplementary Table S4) and searched their protein sequences to see whether they have LIR/AIM. All selected proteins indeed have one or more LIR/AIM, suggesting that they may be ATG8-interating partners.

Apart from the GO analysis, we manually examined the list of potential ATG8-interacting proteins and spotted some very interesting proteins. A manually annotated list is shown in Figure 5E. First of all, like all other ATG8/LC3 interactomes, essential membrane trafficking regulators were identified. These include SH3P2, the ATG8-, and PI3P-binding protein that regulates autophagosome formation; the COPII coat proteins SEC13 and SEC24—COPII vesicles have been revealed to participate in autophagosome formation; EXO70A1, the Exocyst component that had been reported to regulate root growth; Rab2, the Rab GTPase whose mammalian homolog have been reported to regulate autophagy; the Guanosine nucleotide diphosphate dissociation inhibitor GDI1, which is general to Rab GTPases and ROP GTPases; TRS120, the TRAPPII component that is required for cell plate formation; and CLASP, the microtubule-associated protein that regulates both cell division and cell expansion. Plasma membrane- and tonoplast-localized proteins, such as the P-ATPase AHA6, the sodium transporter HKT1, and the Vacuolar H⁺ -Pyrophosphatase AVP1, were also identified. Furthermore, kinase and phosphatases that have been reported to regulate autophagy or stress response were seen in the list. These include the catalytic and regulatory subunits of the plant AMPK, SnRK1/KIN10, the key regulator in autophagy, response to carbon limitation, and circadian clock; Casein kinase 1, an ancient protein kinase that had been reported to regulate selective autophagy; PHOT2, a key blue light receptor and kinase that regulate response to high light; stress responsive kinases ABC1K10A and LECRK-IV.1; TOPP4, an evolutionarily conserved PP1 phosphatase that regulates plant growth and immune response; and ATB’GAMMA, a regulatory subunit of the PP2A phosphatase that is known to balance growth and defense. Other stress response proteins identified include UBP1B, a marker protein for stress granules that stores mRNA during stress conditions, as well as FIERY1/HOS2, SAD2, ADH1, RPM1/RPS3, and PUP1 and PUP3, two purple acid phosphatases that function in response to low phosphate. We concluded that the potential OsATG8a-interacting proteins are worthy of further exploration, and the transgenic lines that we generated can be used by the community in future to identify new OsATG8a-interacting proteins under various developmental and stress conditions.